A Pareto Optimization Solution for a 2nd Order Servo Model Identification

摘要

An ideal servo can be typically modeled as a linear 2nd order dynamic system,rnwhereas a realistic servo is much more complex due to its uncertainties, parasiticrnresonances, non-minimum phase or delay characteristics and non-linearities suchrnas friction and backlash. In cases where the non-linearities are significant, systemrnidentification and evaluation of the friction and backlash components for the servornmodeling is a complicated task. This effort can be simplified if the identification isrnperformed in the frequency domain. However, frequency domain identificationrninvolves simultaneous minimization of a cost functions vector comprised of thernGain and the Phase identification errors. Since in the nonlinear case, the tworncomponents of this cost-functions vector are conflicting, a tradeoff is necessary.rnThe solution to this conflict illustrated herein applies Multiobjective OptimizationrnProblem (MOP) solution in the sense of Pareto optimality. The results of thernoptimization can be applied to enhance servo controller performance or may bernused as an input to a Six-Degree-of-Freedom (6DOF) Monte Carlo simulation of arnservo embedded vehicle system.